This proposal builds on recent advances in using noninvasive electroencephalograhy (EEG) to capture brain activation patterns that can be used to control computers and/or devices. The potential to use such brain-computer (BCI) or brain-machine interfaces (BMI) in physical rehabilitation is now a real possibility with the co-evolution of technologically sophisticated robotic orthoses. Developing such BMI devices for the lower extremites could provide new avenues for restoring mobility function and quality of life among Veterans and others who are disabled after stroke and other neurological injuries, as well as lower limb amputation. In the Baltimore VA's Maryland Exercise and Robotics Center of Excellence, we have developed an impedance controlled ankle robot (anklebot) to enhance paretic ankle motor control and gait function in stroke survivors. Recently we also have shown the first use of noninvasive (EEG) to decode and reconstruct lower extremity joint kinematics for treadmill walking in nondisabled subjects. Our results are very similar to those obtained with direct (invasive) cortical recordings from primates as they performed bipedal treadmill walking. The anklebot provides an ideal test platform to integrate these two lines of research to deveop an innovative BMI for the lower extremity. In this proposal we will validate a novel neural decoding approach for extracting kinematic parameters for ankle movements recorded with the anklebot while using high-density EEG in persons with chronic hemiparesis after subcortical stroke and their age-matched controls. We will also determine the test-retest reliability of the decoded parameters. Both groups of subjects will be trained while wearing the anklebot to generate brain activation patterns to autonomously control the anklebot, demonstrating proof-of-concept that a lower extremity powered orthoses can be operated by using cortical signals. To our knowledge this will be the first demonstration of a lower extremity BMI application. Specifically, this pilot project is designed to test the hypothesis that high density EEG can reliably decode distinct brain activation patterns associated with dorsi-plantarflexion (DF, PF) movements performed in the anklebot. Based on our prior experience using the same approach to train BCI control for the upper extremity, we expect that reconstruction of ankle movements from the EEG signals will yield correlations and signal-to-noise ratios sufficiently high to permit closed loop learning from the decoded EEG, such that the control group and the stroke group will learn to generate specific EEG patterns for controlling the anklebot to move through intended DF-PF movements. This study will provide the first proof of concept that smart prostheses for the lower extremity can be controlled by intention using noninvasive EEG. These results will demonstrate feasibility for our long-range aim of developing a thought-controlled exoskeleton for bipedal locomotion.